resizing – Resizing Layers

Extended Resizing Layers

class neuralnet_pytorch.resizing.Interpolate(size=None, scale_factor=None, mode='bilinear', align_corners=None, input_shape=None)

Down/Upsamples a given multi-channel 1D (temporal), 2D (spatial) or 3D (volumetric) data.

Parameters:
  • size – output spatial sizes. Mutually exclusive with scale_factor.
  • scale_factor – float or tuple of floats. Multiplier for spatial size. Has to match input size if it is a tuple. Mutually exclusive with size.
  • mode – talgorithm used for upsampling: 'nearest', 'linear', 'bilinear', 'bicubic', 'trilinear', and 'area'. Default: 'nearest'.
  • align_corners – if True, the corner pixels of the input and output tensors are aligned, and thus preserving the values at those pixels. If False, the input and output tensors are aligned by the corner points of their corner pixels, and the interpolation uses edge value padding for out-of-boundary values, making this operation independent of input size when scale_factor is kept the same. This only has effect when mode is 'linear', 'bilinear', or 'trilinear'. Default: False.
  • input_shape – shape of the input tensor. Optional.
class neuralnet_pytorch.resizing.AvgPool2d(kernel_size, stride=None, padding=0, ceil_mode=False, count_include_pad=False, input_shape=None)

Applies a 2D average pooling over an input signal composed of several input planes.

Parameters:
  • kernel_size – the size of the window.
  • stride – the stride of the window. Default value is kernel_size.
  • padding – implicit zero padding to be added on both sides.
  • ceil_mode – when True, will use ceil instead of floor to compute the output shape.
  • count_include_pad – when True, will include the zero-padding in the averaging calculation.
  • input_shape – shape of the input image. Optional.
class neuralnet_pytorch.resizing.MaxPool2d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False, input_shape=None)

Applies a 2D max pooling over an input signal composed of several input planes.

Parameters:
  • kernel_size – the size of the window.
  • stride – the stride of the window. Default value is kernel_size.
  • padding – implicit zero padding to be added on both sides.
  • dilation – a parameter that controls the stride of elements in the window.
  • return_indices – if True, will return the max indices along with the outputs. Useful for torch.nn.MaxUnpool2d later.
  • ceil_mode – when True, will use ceil instead of floor to compute the output shape.
  • input_shape – shape of the input image. Optional.

Custom Resizing Layers

class neuralnet_pytorch.resizing.GlobalAvgPool2D(keepdim=False, input_shape=None)

Applies a 2D global average pooling over an input signal composed of several input planes.

Parameters:
  • keepdim (bool) – whether to keep the collapsed dim as (1, 1). Default: False.
  • input_shape – shape of the input image. Optional.
class neuralnet_pytorch.resizing.Cat(dim=1, *modules_or_tensors)

Concatenates the outputs of multiple modules given an input tensor. A subclass of MultiSingleInputModule.

See also

MultiSingleInputModule MultiMultiInputModule Sum SequentialSum ConcurrentSum SequentialCat ConcurrentCat

class neuralnet_pytorch.resizing.ConcurrentCat(dim=1, *modules_or_tensors)

Concatenates the outputs of multiple modules given input tensors. A subclass of MultiMultiInputModule.

See also

MultiSingleInputModule MultiMultiInputModule Sum SequentialSum ConcurrentSum Cat SequentialCat

class neuralnet_pytorch.resizing.SequentialCat(dim=1, *modules_or_tensors)

Concatenates the intermediate outputs of multiple sequential modules given an input tensor. A subclass of Cat.

See also

MultiSingleInputModule MultiMultiInputModule Sum SequentialSum ConcurrentSum Cat ConcurrentCat

class neuralnet_pytorch.resizing.Reshape(shape, input_shape=None)

Reshapes the input tensor to the specified shape.

Parameters:
  • shape – new shape of the tensor. One dim can be set to -1 to let torch automatically calculate the suitable value.
  • input_shape – shape of the input tensor. Optional.
class neuralnet_pytorch.resizing.Flatten(start_dim=0, end_dim=-1, input_shape=None)

Collapses some adjacent dims.

Parameters:
  • start_dim – dim where flattening starts.
  • end_dim – dim where flattening ends.
  • input_shape – shape of the input tensor. Optional.
class neuralnet_pytorch.resizing.DimShuffle(pattern, input_shape=None)

Reorder the dimensions of this variable, optionally inserting broadcasted dimensions. Inspired by Theano’s dimshuffle.

Parameters:
  • pattern – List/tuple of int mixed with ‘x’ for broadcastable dimensions.
  • input_shape – shape of the input tensor. Optional.